Thesis. The genuine inflection in behind-the-meter power — revenue up 165.5%, GAAP operating income and operating cash flow both turned positive, a $20B backlog, and every major US hyperscaler having validated the product.
Reviewed 19 Sep 2026, after Q2 FY2026 results on 28 Jul and the S&P 500 inclusion announced 4 Sep. Next review after Q3 FY2026 earnings on 5 Nov 2026.
What it does
Bloom makes solid oxide fuel cells for on-site power. A fuel cell converts fuel directly into electricity through a chemical reaction rather than by burning it — no combustion, no moving parts, very low emissions, quiet. It runs on natural gas or hydrogen.
The reason datacentres want them has nothing to do with emissions. They can be installed in months rather than years, on site, without waiting for grid interconnection. Speed-to-power is the product. That is also the whole fragility of the thesis: Bloom monetises a scarcity, and scarcity is a thing other people are actively trying to remove.
Fuel cells cost more per kilowatt than gas turbines. Customers pay the premium for time, not for electricity — so the value of the product falls as turbine lead times fall.
Bull case
- Q2 FY2026 revenue $1.07B, +165.5% YoY against $827M consensus — a very large beat. Product revenue $935.4M, +215.4%, driven by hyperscalers, neoclouds and AI labs
- Non-GAAP EPS $0.78 against ~$0.41 consensus — a fourth consecutive beat
- Swung to $182.2M of GAAP operating income from a small loss a year earlier, and operating cash flow flipped positive to $226.4M. Both profitability tests passed in the same quarter
- FY26 guidance raised to $3.9–4.2B revenue (~100% growth) and non-GAAP EPS $2.55–2.85
- ~$20B backlog, plus a $5B Brookfield AI infrastructure partnership
- CEO KR Sridhar: "all the major US hyperscalers and over a dozen US neoclouds, AI labs, and colocation data center operators have validated and approved our power solutions for their AI factories"
- NBIS selected Bloom fuel cells for its 300 MW Vineland, New Jersey datacentre — the first publicly named neocloud win. Stock rose 13%
- Added to the S&P 500 effective at the open on 21 September 2026, forcing passive buying and reflecting the scale the business has reached
Bear case
- The scarcity Bloom monetises is under direct attack. SpaceX's Bastrop turbine-blade foundry targets up to 18 months faster turbine availability, and is scaling toward 30 GW of output by 2030. Neither is immediate — the pressure is dated 2028+ — but both point the same way
- Fuel cells still cost more per kW than gas turbines. The value proposition is speed, not price, and it evaporates if the queue shortens
- Extremely volatile — the stock gained and gave back 11% on the earnings day itself, with large drawdowns inside a strong year. High beta cuts both ways around the index inclusion
- Dependent on natural gas prices and availability
- Competes with FuelCell Energy (weaker) and increasingly with fast-deploy turbines and gensets from and CAT
- Hyperscaler validation is management's characterisation; no major hyperscaler is individually named
Major customers
- All major US hyperscalers — validated and approved per management, but not individually named
- Over a dozen US neoclouds, AI labs and colocation datacentre operators
- NBIS (Nebius) — 300 MW Vineland, New Jersey. The first publicly named neocloud win
- Brookfield — $5B AI infrastructure partnership
- Commercial and industrial customers for the legacy stationary power business
Read-through. Bloom is the clearest beneficiary of grid interconnection scarcity in this file, and neocloud capacity announcements are now a leading indicator for its orders — note that the Nebius win was disclosed on Nebius's call, not Bloom's, so watching NBIS, CRWV and IREN gives an earlier read than watching Bloom. It sits opposite and CAT in the same customer decision — turbine, reciprocating engine or fuel cell — so those three move against each other on the same news. Anything that shortens turbine lead times compresses Bloom's advantage directly.
What would change the view
- Whether ~100% growth holds for another two quarters
- Gross margin by customer type, and cost per kW against gas turbines — the economics behind the speed premium
- Additional named hyperscaler or neocloud wins
- Gas turbine lead times industry-wide — the single variable that erodes the thesis
- Backlog conversion from the ~$20B base
- Brookfield partnership deployment pace
- How the stock behaves after the 21 September index inclusion, once the forced buying is done
Update log
19 Sep 2026 — Converted to the standard template. Flagged an internal contradiction: the My Verdict property reads Conviction while the 7 September note on this page said "Verdict unchanged: Constructive." One of the two is wrong and it should be settled before this file is shown to anyone.
4 Sep 2026 — Added to the S&P 500, effective at the open on 21 September, alongside Everpure and Illumina. Stock rose ~6% after hours.
29 Aug 2026 — The SpaceX turbine-blade foundry story is the clearest threat to the speed-to-power advantage; GE Vernova's capacity expansion toward 30 GW by 2030 points the same way. Neither is immediate.
28 Jul 2026 — Q2 FY2026 reported. Revenue $1.07B (+165.5%) against $827M consensus; product revenue $935.4M (+215.4%); non-GAAP EPS $0.78 against ~$0.41. GAAP operating income $182.2M and operating cash flow $226.4M both turned positive. FY26 guidance raised to $3.9–4.2B revenue and $2.55–2.85 non-GAAP EPS.
Research and education only — not investment advice.
Made with Bullet